Save Your Mobile Payments; This AI Can Spot a Fake Review

September 20, 2018         By: Steven Anderson

Whether we’re looking to buy something, go somewhere, or any combination of the preceding, we often turn to reviews first to figure out just what it is we’re picking up or where it is we’re going. Yelp has made the restaurant review a must-check, even if it’s becoming a bit insufferable as evidenced by the South Park song about Yelpers. This in turn makes fake reviews a particular hazard, and one that some new reports sent our way suggest can be spotted and fought against.

The problem, as defined by Aalto University doctoral student Mika Juuti, is that fake reviews built around algorithms are actually easy to build, fast to generate, and sound for all intents and purposes like they were written by a reasonably astute adult human, which as we know, is about 70 to 80 percent of all adult humans. Juuti notes that such reviews can not only artificially create a positive brand image, but can also be weaponized against the competition, creating a negative brand image at the competitor’s doorstep.

A University of Chicago study from 2017, meanwhile, started spotting issues in reviews that could help spot potential fakes. While humans had a tough time at this task—Juuti notes that up to 60 percent of the fake reviews were considered real by human reviewers—the AI classifier Juuti and a team developed had much better luck. The classifier was actually part of a larger project at Juuti’s Aalto University known as “Deception Detection via Text Analysis.”

Essentially, what we have is potentially the makings of a tool—which could likely go mobile besides—that can spot fake reviews generated by AI and point these out. That could be a huge help when it comes to figuring out where to eat, and where to avoid. Of course, the problem with this tool is that it looks for fake reviews generated by AI. Fake reviews generated by humans would still be possible, so that makes the Aalto University tool less than useful.

Still, a fake review spotted and removed is one less fake to contend with, and that will help mobile payments users, mobile shoppers, and those who check their mobile devices for possible restaurant choices all over the world.